GTC 2025: AI Trends & Strategic Implications for FIT

Sizhe Liu

USRD, FIT

AI Compute Knows No Rest

  • Massive Scale: AI demands unprecedented data throughput & compute (10k-100k+ GPUs).

Huge scale spike
  • Infrastructure is Key: Data centers become “AI Factories” needing ultra-high-speed, low-latency interconnects.

Data center -> AI factory
  • AI Application Booming: Among 2900+ GTC sessions, only ~180 focused directly on compute hardware

Application is the mainstream

Key Trend 1: Hardware & Infrastructure Scale

  • Massive GPU Clusters: 50k-100k+ GPUs drive demand.
  • High-Performance Fabric: 800G+ networks (IB/Ethernet), >95% utilization needed.
  • Evolving Standards: PCIe 6.0/7.0 & CXL crucial for disaggregation/scale-up.

LLM compute and communication profiling showing bursty communication patterns

Key Trend 2: AI Transforming Design & Manufacturing

  • AI-Driven Design: Faster simulation, optimization, generation.
  • Smart Manufacturing: Enhanced QC, process optimization (Digital Twins), predictive maintenance.
  • Human-Robot Collaboration: Intuitive interaction, flexible automation.
  • FIT’s Opportunity: Leverage AI internally (Design, Mfg, QC) & explore customer solutions.

YooDriveCloud Data Closed-loop Toolchain

Key Trend 3: Specialized AI Domains Emerge

AI Factories: * Heterogeneous compute * Extreme power/thermal needs * Rack-scale interconnects (DACs/Optics)

Industrial Robotics: * Ruggedized, low-latency, deterministic * Harsh environments (TSN)

Autonomous Vehicles: * Safety-critical, high-bandwidth * Resilient, compact

FIT’s Strategy: Develop domain-optimized interconnect families for these high-value markets.

Industrial robotics requires specialized interconnects

Edge AI systems for autonomous vehicles

Evidence: Market Growth and Performace Gains

Figure 1: Projected AI Infra & Interconnect Market Growth
  • Significant growth projected in AI server spending & high-speed interconnects
  • Specialized domains (Industrial, Robotics, AV) show rapid growth
(a)
(b)
Figure 2: AI-Driven Mfg. Quality & Efficiency Gains
  • AI offers significant potential for improving Mfg. Quality, Efficiency, & Time-to-Market (relevant for FIT & customers).

Strategic Imperatives for FIT

  1. Lead High-Speed: Drive PCIe 6/7, CXL, 112G+ SerDes. Maximize CDFP/OSFP-XD.
  2. Enable Architectures: Solutions for disaggregation & scale-up (w/ partners like Alphawave).
  3. Specialize for AI: Optimize for bursty traffic, low latency, reliability.
  4. Target Domains: Tailored solutions for AI Factories, Industrial, Automotive.
  5. Manufacturing Excellence: Leverage AI in 3 directions: precision, quality, scale.

Conclusion: FIT Enabling the AI Future

  • The AI revolution hinges on high-performance interconnects – FIT’s core strength.
  • Market trends (scale, specialization, new standards) align perfectly with FIT’s capabilities and recent successes (PCIe 6/7, CDFP/OSFP-XD).
  • Strategic focus on high-speed leadership, domain specialization, and manufacturing excellence will solidify FIT’s position as a critical enabler of AI infrastructure.
  • FIT is uniquely positioned to connect the future of AI.